SUPPLEMENTARY INFORMATION:

Accurate CNS tumor classification can be challenging when tumors show overlapping histology, limited tissue or atypical features. A meaningful fraction of cases remains unclassified or assigned with limited confidence with current molecular tools. These diagnostic ambiguities directly affect subtype and grade assignment which, in turn, influence treatment planning, prognosis, and clinical trial eligibility. Variability across observers and institutions can lead to additional testing, delays, and inconsistent diagnoses.

The NCI/Bethesda classifier addresses this gap using DNA methylation patterns as a robust molecular fingerprint. It applies a stratified machine learning framework to extend diagnostic coverage and improve assignment confidence for CNS tumors. The classifier was developed from a rigorously curated reference set of over 16k methylation profiles, structured into 22 tumor families and 133 clinically relevant diagnostic classes and includes 21 recently developed methylation classes not represented in existing tools. The approach has been validated across multiple independent cohorts (n = 5,875) and supports deployment through MethylScape, a public web-based portal that streamlines classifier execution for broad accessibility. In an NIH validation cohort analysis of 1,204 high-confidence matches with pre-methylation diagnoses available, methylation profiling confirmed the initial diagnosis in 25.6% of cases while driving clinically meaningful diagnostic evolution in the remainder, including refined diagnosis (subtyping) in 14.6%, new diagnosis with increased precision in 54.7%, and substantial diagnostic reclassification in 5.0%—changes that are expected to affect patient management in the reclassification subset.

Renal neoplasms present a parallel diagnostic challenge due to morphologic and molecular heterogeneity, overlapping microscopic features, and interobserver variability. As a result, a subset of cases are unclassifiable even after immunohistochemical, mutation and cytogenetic workups. To address this, the Kidney Classifier component of this platform leverages genome-wide DNA methylation profiling (feasible on formalin-fixed paraffin-embedded tissue using robust array-based methods). It was developed through examination of methylation signatures from over 2,000 renal neoplasms, identifying 23 coherent methylation groups that correlate with known tumor types and reveal clinically relevant novel subtypes. A machine learning classifier trained on 1,284 samples was externally tested on 287 renal neoplasms, demonstrating >90% concordance between expected neoplasm type and high-score methylation-based classification, with discordant cases highlighting opportunities for diagnostic reclassification and improved precision in challenging renal tumor evaluations.

Licensing and collaboration opportunities include commercial development of methylation-based diagnostic tests and/or software-enabled classification solutions for clinical laboratories, reference labs, and diagnostic companies. Partners may engage in external validation (retrospective and prospective), assay standardization, integration into pathology workflows and reporting systems, and extension of classifier coverage to additional tumor types and multi-institutional datasets. Consistent with consensus recommendations for complementary classifiers, the inventors are also interested in collaborations that: (1) operationalize multi-classifier strategies (concordant/complementary prediction to increase confidence, (2) leverage discordance to trigger orthogonal follow-up) and (3) accelerate clinical translation through scalable deployment models- including CLIA laboratory workflows and regulated diagnostic pathways.

“This Notice is in accordance with 37 CFR 404.4 Authority to grant licenses.”

>NIH Reference Number: E-105-2023-0.

Related Technologies: N/A.

Product Type: Oncology.

Therapeutic Area(s): Diagnostic.

Development Stage: Prototype.

Publications:

• Aldape K., et.al. cIMPACT-NOW update 9: Recommendations on utilization of genome-wide DNA methylation profiling for central nervous system tumor diagnostics. Neurooncol Adv. 2025 Jan 3;7(1):vdae228. PMID: 39902391.

Patents: PCT/US2024/054179, filed November 1,2024.

Potential Commercial Applications:

• Molecular classification and diagnosis of CNS and kidney tumors, including difficult-to-classify and low-confidence cases.
• Diagnostic subtyping aligned to WHO-guided entities.
• Reference-lab and hospital-lab deployment.
• Clinical trial stratification and translational research cohort harmonization.
• Multi-classifier diagnostic decision support.
• Cancer treatment development.

Competitive Advantages:

• Developed from a rigorously curated, large reference set of methylation profiles.
• Clinically relevant CNS diagnostic.
• CNS tumor methylation classes not represented in existing tools, expanding diagnostic coverage.
• Clinical-impact analysis creating superior diagnostic precision.
• Superior classifier for kidney cancer.
• Superior classifiers for multiple solid, difficult-to-diagnose tumors.
• Integrative into clinical workflows, improving diagnostic practices and enhancing patient care.

Collaboration Opportunity: Researchers at the NCI seek licensing and/or co-development research collaborations to further develop and possibly expand the classification capabilities of the software.